The Food That Never Gets Scanned

In August, USDA published Food-At-Home Purchases by Supplemental Nutrition Assistance Program (SNAP) Households in 2023. It describes what households buy at the grocery store, sorted by department and aisle, comparing households that used SNAP against households that did not. If you work anywhere near food security, somebody is going to send it to you this fall with an opinion attached.

The report itself does not carry an opinion, and that restraint is worth borrowing. USDA writes that "differences observed in the aggregate expenditure shares, even those that are statistically significant, should not be interpreted as establishing or implying any causal relationship between SNAP participation and household food expenditure patterns." Keep that sentence handy; it tends to fall off as a report travels.

What I want to do here is what we did with Map the Meal Gap last month: be plain about what a national dataset is, what it is good for, and what it structurally cannot see. The limits are not failures of effort. They come from the design, and the design is the reason the study can cover the whole country in the first place.

What the study is

The analysis draws on 51,891 households in the Circana Consumer Network Panel, who log their grocery purchases with handheld scanners and a phone app. Together they reported roughly $875 million in food-at-home spending during calendar year 2023. Of those households, 7,307 were counted as SNAP households because they reported paying with SNAP benefits at least once that year. Purchases were sorted by retail category rather than nutrition, so the units are departments and aisles: produce, dairy, frozen, beverages, and so on.

USDA is direct about who ended up in the sample. The panel's SNAP households "have higher incomes and meaningfully different household composition (e.g., fewer households that include an elderly person, fewer single-adult households) than the SNAP population," and so are "likely to receive less in SNAP benefits than the SNAP population on average." The report describes a set of households that used SNAP at some point in 2023. It is not a portrait of SNAP participants as a group, and USDA does not present it as one.

The line most people will quote is the produce department, where SNAP households in the panel put 14.0 percent of food-at-home spending against 19.4 percent for non-SNAP households. Read it next to the causation sentence above, and then next to the two things the study cannot see at all.

Two things it cannot see

Food that was never purchased. A consumer panel records transactions. A pantry bag is not a transaction. Neither is a produce prescription box, a senior box, a backpack sent home on a Friday, a second-Saturday church distribution, or a pallet of milk moved on a food recovery run before it turned. None of it shows up in scanner data, because none of it was bought. For the households most likely to be using both the grocery store and the charitable system in the same week, the panel sees one side of the ledger and has no way of knowing what sat on the other.

Nobody can size that other side either. The national food bank networks publish distribution totals for the organizations inside their networks, and those totals are useful. They do not include the church basement that opens two mornings a week, the volunteer-run closet behind the fire station, or the mutual aid group that formed after the storm and never registered with anyone. Verifying a single county turns up plenty of places like that, answering to no reporting structure. There is no national number that includes them, because nobody has the county lists to add up.

What was reachable. An expenditure share is the last step in a long chain: which stores are within driving distance, whether the car ran that month, the bus schedule, the hours somebody could actually shop, what the produce section looked like on the fourteenth versus the twenty-eighth, whether the store that closed in 2022 was ever replaced. A grocery basket records the end of the chain and nothing about the chain. No national panel can see any of that, and I do not think one should be asked to. It is a county-level question, and it gets answered at the county level, by people who go look.

The purchase record and the local picture

Here is how I would put the new report to work.

Use the report for what it is: a careful national look at grocery purchasing, with the caveats printed in the document itself. It tells you something real about one channel through which food reaches households. It does not try to explain why, and says so.

Then set it beside the thing it was never built to include. Where is food being given away in your county, by whom, on what days, at what hours, and has anybody confirmed that this month. Where a family that runs out of groceries on the twenty-sixth actually has somewhere to go. Which corridors have a produce-carrying grocery store within a reasonable trip, and which do not.

When a national report starts circulating, the most useful question a funder can ask is a quiet one: what does the current picture in our own county show? That is an answerable question. It just takes somebody walking it.

That second half is what FastRoots builds: every food provider in a county, confirmed through real conversations by paid local stewards, verified every quarter, crossed with tract-level need. The national report can describe the transaction. The living map covers the rest of the way food gets to a table.

If you want to see what that looks like for your region, start here.